Irrational Capital Deep Fin Demo

Case Study 1: Understanding the Inputs to the Human Capital Factor

In this Case study we're going to better understand what the Human Capital Factor is by exploring the input data and calculating for a small sample of companies.

Key takeaways:

Part 0: Set up

Begin by loading all packages used and defining the company universe for this exercise - which will be the same companies from the survey.

Part 1: Proprietary Data

With our notebook set up we'll now explore the data that is provided by our private data partner. We'll begin by analyzing the data exactly as it is provided by the source (our private data partner).

Group the various columns into categories

Look at an example row to better understand what the data looks like.

Now we'll run our process that simplifies these flags down to the key information

With an understanding of our data as provided by the data partner, it's now time to run the remaining cleaning steps to get the data ready to by loaded by the model

With the clean data written to the data lake, now we'll re-load it using our loader function

Private Data Exploration

The private data is now loaded exactly as it is needed to calculate the Human Capital Factor Score. With our data analysis-ready, we can do some exploration to see what exactly is in this data.

The proprietary data also has a wealth of demographic information. Each respondent is identified by their tenure, salary, management level (job grade), full-time status, and gender.

Part 2: Public Data

For part 2 we encourage interested participants to follow along and do their own exploration using the provided participant notebook.

In this section we'll get a better understanding of the public data then we'll look at financial returns for a naive strategy using the public data.

Begin by loading the analysis-ready data using our loader functions.

Participant notebook starts here:

Now we'll expore the public data in similar way to what we did with the private data

Incorporate Returns

Next we'll bring in the return stream for the public data and test a naive strategy based on this data.

Part 3: HCF Process

With all of our analysis-ready data loaded the final step is to run the HCF process and look at the results. For simplicity, we'll keep the process to the the 36 companies in from our earlier survey.